Skip to content

Question about poisson GLM

3 messages · rosanna di maggio, Bob OHara, Rob Robinson

Original
rosanna di maggio
# 
Dear all,

I'm trying to perform a poisson GLM with lme4 using the number of
individuals as dependent variable and i) years of study and ii) different
study areas as factors... So far so good..

But the results are quite curious...  The estimates and standard errors are
huge.. Is is normal? I don't think so...


This is an example of what happen:


Call:
glm(formula = NTOT ~ areaN * anno, family = poisson)

Deviance Residuals:
    Min       1Q   Median       3Q      Max
-3.7465  -1.0629  -0.5000   0.3132   6.8188

Coefficients:
              Estimate Std. Error z value Pr(>|z|)
(Intercept) 140.786646  27.766187   5.070 3.97e-07 ***
areaN        71.479294  11.876992   6.018 1.76e-09 ***
anno         -0.069199   0.013823  -5.006 5.56e-07 ***
areaN:anno   -0.035572   0.005912  -6.017 1.78e-09 ***




I cannot understand the reason of these results... Someone can help me?



Thank you

Rox
Bob OHara
#
On 27/10/14 15:38, rosanna di maggio wrote:
It's difficult to say without seeing that data, but a couple of things 
spring to mind:

1. If the covariates are meant to be continuous, then did you centre and 
scale them?
2. If they are meant to be factors, did you tell R this? e.g. anno <- 
factor(anno).

Bob

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Rob Robinson
    

    

      
      #
    

  


  

      
Rox
 If anno is year and year is 2000:2010 (or whatever) and you have a
negative relationship then extending this back to 0 (the intercept) will
yield quite a large number (2000*0.07=140 for instance). Does coding year
as year of study (1:10) or even (-4:5), if you wanted the intercept to be
in the middle of your study period, help?
cheers
rob


******* View the latest ringing report at www.bto.org/ringing-report
************

Dr Rob Robinson, Principal Ecologist
British Trust for Ornithology, The Nunnery, Thetford, Norfolk, IP24 2PU
Ph: +44 (0)1842 750050       T: @btorobrob
E: rob.robinson at bto.org      W: www.bto.org/about-bto/our-staff/rob-robinson

====== "How can anyone be enlightened, when truth is so poorly lit" =======


On 27 October 2014 14:38, rosanna di maggio <rosannadimaggio at gmail.com>
wrote: